Research map / Image generation / Fast diffusion sampling
Unifying Deep Stochastic Processes for Image Enhancement
Wojciech Maciej Kozłowski, Radosław Kuczbański, Kamil Adamczewski, Karol Szczypkowski, Maciej Zieba
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use unconditional diffusion (DM-VP) instead of conditional InDI | InDI · Low-light Enhancement (LOL) | PSNR: 21.48 -> 23.01 | Improved |
| use unconditional diffusion (DM-VP) instead of ResShift | ResShift · Facial Super-Resolution (FFHQ x8) | PSNR: 27.66 -> 27.86 | Improved |
| use ResShift instead of DM-VP | DM-VP · General Super-Resolution (DIV2K) | PSNR: 24.12 -> 24.36 | Improved |
| use diffusion bridge BBDM instead of DM-VP | DM-VP · Deraining (Rain1400) | PSNR: 29.02 -> 30.10 | Improved |
| use diffusion bridge I2SB instead of DM-VP | DM-VP · Colorization (ImageNet) | PSNR: 21.73 -> 22.39 | Improved |
| lower ResShift temperature to match InDI default tau=0.06 (ablation) | ResShift · image enhancement tasks | quality | Hurt |
| add endpoint regularization (UniDB) over GOUB (ablation) | GOUB · Facial Super-Resolution (FFHQ x8) | PSNR: 27.24 -> 27.59 | Improved |
| add endpoint regularization (UniDB) over GOUB (ablation) | GOUB · Deraining (Rain1400) | PSNR: 28.38 -> 28.02 | Hurt |
| use deterministic samplers with diffusion bridges instead of stochastic ones (ablation) | diffusion bridges · image enhancement tasks | LPIPS | Hurt |
| change discretization technique (ablation) | all methods · image enhancement tasks | sample quality | No clear effect |
| use ancestral sampling instead of other samplers (ablation) | all methods · image enhancement tasks | sample quality | Improved |
Closest papers by meaning
- Warm Diffusion: Recipe for Blur-Noise Mixture Diffusion Models (ICLR 2025 · similarity 0.87)
- UniDB: A Unified Diffusion Bridge Framework via Stochastic Optimal Control (ICML 2025 · similarity 0.86)
- The Blessing of Randomness: SDE Beats ODE in General Diffusion-based Image Editing (ICLR 2024 · similarity 0.86)
- Denoising Diffusion Bridge Models (ICLR 2024 · similarity 0.86)
- Spectral Progressive Diffusion for Efficient Image and Video Generation (NeurIPS 2026 · similarity 0.86)
- Exploring the Design Space of Diffusion Bridge Models (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
